Funnel Step

Origin & evolution

The funnel as a shape comes from advertising. Elias St. Elmo Lewis sketched the stages of a sale around 1898, and William W. Townsend married that progression to the funnel image in Bond Salesmanship (1924), describing the work as "the forcing by compression of a broad and general concept of facts through a funnel" (Purchase funnel, Wikipedia). For most of the next century the steps were marketing abstractions: awareness, interest, desire, action.

Digital products made each step a counted event. The reframing that growth teams use today came from Dave McClure, who presented "Startup Metrics🔢MetricStrategyA unified metric that measures progress, health, or behaviour across the productView reference → for Pirates" at Seattle Ignite in 2007 and named five stages, acquisition, activation, retention, referral, and revenue, that spell AARRR (Inc.). Each pirate-metric stage is a funnel step with its own instrumentation, and the discipline McClure pushed was to track step conversion against vanity totals.

The thinking has since loosened the strict ordering. Growth practitioners now treat a funnel step as one node a user can enter, skip, or revisit, and many products favour retention-first reorderings such as RARRA over a rigid acquisition-down sequence (Mind the Product). The step survives as the unit; the path between steps is what teams keep arguing about.

How it works in practice

A B2B SaaS team maps signup-to-activation as four steps: account created, workspace set up, first project imported, first teammate invited. Over a month, 1,000 accounts are created, 740 set up a workspace, 300 import a project, and 280 invite a teammate. Aggregate signup-to-activation reads 28 percent, which tells the team almost nothing about where to act.

Stepwise, the leak is obvious. The setup step holds 74 percent, but the import step drops to 41 percent (300 of 740). That single transition loses more users than the other three combined. The team ships a sample-data importer, the import step rises to 62 percent the following month, and downstream invites climb without anyone touching the invite flow. The funnel step located the wound; the aggregate rate only confirmed there was bleeding.

Funnel step vs. its neighbours

• Funnel is the ordered whole; a funnel step is one stage inside it. The funnel owns the overall conversion question; the step owns a single transition rate and its drop-off. A funnel⬇️FunnelGrowthA conversion funnel tracking user progressionView reference → without enumerated steps cannot be diagnosed, only lamented.
• Customer journey stage🚏Customer Journey StageCustomer SuccessA stage in the post-sale customer journeyView reference → is qualitative and experience-led, spanning touchpoints👆TouchpointCustomer SuccessA customer interaction touchpointView reference → over weeks (onboarding, adoption, advocacy). A funnel step is a quantitative, instrumented checkpoint with a conversion percentage attached. A journey stage may contain several funnel steps.
• Event schema🔠Event SchemaData & AnalyticsAn event schema for trackingView reference → is the definition of the tracked behaviour, the named event and its properties. A funnel step is measured by events; the schema is what makes the step countable. Change the schema and every step reading it shifts.

In the graph

In the Unified Product Graph, a funnel step sits in the Business, GTM & Growth region as a leaf under its parent funnel, joined by funnel_contains_funnel_stepFunnelcontainsFunnel Stephierarchy. Three edges give it analytic teeth: funnel_step_reveals_needFunnel SteprevealsNeedcross-domain ties a drop-off back to an unmet user need🫀NeedUserA user need, pain, desire, or constraintView reference →, while funnel_step_tracks_event_schemaFunnel SteptracksEvent Schemacross-domain and its inverse event_schema_tracks_funnel_stepEvent SchematracksFunnel Stepcross-domain bind the step to the instrumentation that quantifies it. That binding is the point of modelling the step at all: a funnel step with no event schema is a claim no one can verify, and the graph makes that absence queryable rather than invisible.